📣 Releasing r37

[joelim-work]

It hasn't even been a month since the last release, but there has been so many changes (including breaking changes) recently that I should probably make a new one, otherwise it will become too overwhelming for users that are wanting to upgrade.

Below are the release notes I am planning to include:

## [r37](https://github.com/gokcehan/lf/releases/tag/r37)

### Changed

- The default paths of files read by `lf` is changed on Windows, to separate configuration files from data files (#2051).
  - Configuration files (`lfrc`/`colors`/`icons`) are now stored in `%APPDATA%`, which can be overridden by `%LF_CONFIG_HOME%`.
  - Data files (`files`/`marks`/`tags`/`history`) are now stored in `%LOCALAPPDATA%`, which can be overridden by `%LF_DATA_HOME%`.
- The change for following symbolic links when tagging files from the previous release has been reverted (#2055). The previous change made it impossible to tag symbolic links separately from their targets, and also caused `lf` to run slowly in some cases.
- The existing `globfilter` and `globsearch` options are now deprecated in favor of the new `filtermethod` and `searchmethod` options, which support regex patterns (#2058).
  - `set globfilter true` should be replaced by `set filtermethod glob`.
  - `set globsearch true` should be replaced by `set searchmethod glob`.
- File sizes are now displayed using binary units (e.g. `1.0K` means 1024 bytes, not 1000 bytes) (#2062). The maximum width for displaying the file size has been increased from four to five characters.

### Added

- `dircounts` are now respected when sorting by size (#2025).
- The `info` and `sortby` options now support `btime` (file creation time) (#2042). This depends on support for file creation times from the underlying system.
- The selection in Visual mode now follows wrapping when `wrapscan`/`wrapscroll` is enabled (#2056).
- Input pasted from the terminal is now ignored while in Normal mode (#2059). This prevents pasted content from being treated as keybindings, which can result in dangerous unintended behavior.
- The Command-line mode completion now supports keywords for the `selmode` and `sortby` options (#2061), as well as the `info` and `preserve` options (#2071).
- Command line flags are now exported as environment variables in the form `lf_flag_{flag}` (#2079).
- Support is added for terminal escape sequences that disable text styles (#2101).

### Fixed

- `dircounts` are now automatically populated after enabling it (#2049).
- A bug where directories are unsorted after reloading when `dircache` is disabled is now fixed (#2050).
- Filenames are now escaped when completing shell commands (#2071).
- A bug where completion menu entries are misaligned when containing fullwidth characters is now fixed (#2071).
- The `on-load` command now passes all files in the directory as arguments, not just files visible to the user (#2077).
- Failure to move files across different filesystems is now shown as an error instead of a success in the UI (#2085).
- Errors are now logged correctly when there are multiple errors during move/copy operations (#2089).
- The progress for copy operations is now displayed immediately in the UI, even if it takes time to calculate the total size of files to be copied (#2093).

In addition to the changes above, there has also been various refactors of the codebase since (including a rewrite of the completion functionality in complete.go).

---

Given the magnitude of the changes, I think it is worth waiting at least a few weeks to see if there are any regressions, before making the actual release. I am unlikely to merge new changes during this time. If anyone has any queries or concerns, please let me know.

[joelim-work]

I have decided to add #2101, it is a trivial patch which I doubt will break anything.

[CatsDeservePets]

Interesting! I did not know that there are escape sequences to disable specific attributes. I only knew about "\033[0m" to reset them completely.

[joelim-work]

I only found out about it recently myself. It seemed strange to not add it in, given that it is standardized and also simple to implement. And maybe someone will find it useful, e.g. disable bold without also disabling yellow foreground on blue background.